Maya Longpre-Croteau1, Joël Bêty1, Valérie Bertrand3, and Marc-André Villard2
1 Département de biologie, chimie et géographie, Université du Québec à Rimouski, 300 allée des Ursulines, Rimouski, QC, Canada, G5L 3A1
2 Department of Biology, Mount Allison University, 63B York St, Sackville, NB, Canada, E4L 1E2
3 Département de biologie, Université de Moncton, 18 Antonine-Maillet Ave, Moncton, NB, Canada, E1A 3E9
1.3 ABSTRACT
The selection of a breeding territory by an individual is a critical step in its life history, that can have major fitness consequences. In some species, individuals have been shown to undergo a “prospection” period, during which they collect information to increase their likelihood of selecting high-quality breeding sites. In this study, we investigated whether landscape structure can influence the ability of individuals to visit potential breeding territories during the dispersal period. Our focal species, the Ovenbird (Seiurus aurocapilla), has been shown to move more readily across the landscape when the proportion of mature deciduous forest cover increases moves less readily when the proportion of conifer plantations in the landscape increase. We expected that the probability of putative prospectors in potential breeding habitat patches would decrease with forest management intensity in the surrounding landscape. In each study plot (n=6), we captured, and colour banded territorial males in June. Then, we used constant-effort mist netting to capture individuals during the dispersal period, in late July and early August. We captured a total of 77 individuals deemed to be prospectors, i.e. unbanded individuals that likely came from other plots. We then modelled the effects of landscape metrics at three spatial scales on capture rates of the putative prospectors. Capture rates were negatively related to the proportion of conifer plantations at the finest scale (500 m radius), and to the mean distance between patches deemed permeable to Ovenbird movements at all spatial scales. These results suggest that
both matrix composition and habitat configuration can influence Ovenbird prospecting movements. Our study thus suggests that intensive forest management can reduce access to potential breeding sites and, in turn, the overall productivity of such populations.
Keywords: Landscape structure, Prospecting, Fragmentation, Connectivity, Dispersal, Forest management, Ovenbird, Seiurus aurocapilla
1.4 INTRODUCTION
Agriculture, forest management, and urbanization are major causes of habitat loss and fragmentation, which in turn are among the main causes of biodiversity loss worldwide (Haddad et al. 2015). Current trends in land use predict a continued increase in habitat loss and fragmentation across most of the world’s biomes (Newbold et al. 2015). Habitat fragmentation, the process by which habitat is broken apart by the creation of new land cover types, is considered harmful to a wide range of species, especially to ecological specialists (Carrara et al. 2015). The negative influence of habitat fragmentation on population size and overall biodiversity is becoming increasingly clear (Wiegand et al. 2005, Haddad et al. 2015, 2017). However, a recent paper (Fahrig 2017) has fuelled the debate on the relative influence of habitat loss versus fragmentation. While habitat loss is a recognized driving factor in the decline of global biodiversity, the importance of habitat fragmentation cannot be ruled out based on both theoretical and empirical evidence (Haddad et al. 2017, Fletcher et al. 2018a).
For instance, habitat fragmentation can negatively affect population viability through a reduction in functional connectivity (Aben et al. 2012, Vasudev et al. 2015), which represents a species’ behavioural response to landscape structure (Taylor et al. 2006, Baguette et al. 2013). Functional connectivity reflects the relative ease with which individuals can move across the landscape. Hence, high functional connectivity increases the likelihood that individuals will detect and colonize small or relatively isolated habitat fragments (Matthysen and Currie 1996, Kristan 2006) and it is critical for the persistence of those populations (Haché and Villard 2010, Pavlacky et al. 2012, Duarte et al. 2016).
Because the movements of small organisms remain difficult to track with a good resolution over long periods (but see Ponchon et al. 2013 and Hallworth and Marra 2015), various indirect empirical approaches have been used to estimate functional connectivity,
such as translocation (Bélisle et al. 2001, Volpe et al. 2014, Betts et al. 2015) and gap crossing experiments (Desrochers and Hannon 1997, Bowman and Fahrig 2002). Such indirect approaches have generally been applied on adults, as it is easier to manipulate their motivation to move, e.g. through territorial challenges (gap crossing) or by translocating territorial individuals to compare their homing movements in different landscape structures. While Volpe et al. (2014) have reported that movement parameters were similar between post-translocation and daily movements in the Green Hermit (Phaethornis guy), these approaches may not accurately reflect natal dispersal movements because the motivation to move is different and individuals may not have the same knowledge of their landscape. In addition to these experimental approaches, many modelling programs now exist that help identify possible dispersal trends, these, however, require baseline information that must first be obtained in the field (circuitscape, linkagemapper). Genetic markers have also been used to assess the influence of landscape structure on dispersal (Callens et al. 2011, Pavlacky et al. 2012), but this approach is only relevant to larger spatial and temporal scales (Vandergast et al. 2019).
In birds, natal dispersal generally represents the longest movement an individual will perform over its lifetime, outside of migration (Greenwood and Harvey 1982). Additionally, it is difficult to characterize because the individual’s destination is yet unknown. However, natal dispersal plays a crucial role in population dynamics and gene flow, as individuals become relatively site-faithful after their first successful reproduction in many species (Greenwood and Harvey 1982, Pärt and Gustafsson 1989).
In some species, individuals have been shown to gather information on potential breeding sites before selecting a territory (Reed et al. 1999, Doligez et al. 2004, Thériault et al. 2012). Gathering information on site quality gives individuals an advantage when the time comes to settle on a territory (Reed et al. 1999, Doligez et al. 2004, Ponchon et al. 2013), especially in migratory species, where younger individuals arrive later than more experienced
ones (Porneluzi et al. 2011, Sherry et al. 2016). Prospecting may provide information on habitat quality, food availability, and the location and breeding success of conspecific or even other species, leading to better-informed territory selection (Reed et al. 1999, Thomson et al. 2003, Nocera et al. 2006). Prospecting is often performed by dispersing individuals during or immediately after the breeding season, when social information and environmental cues are likely most available (Reed et al. 1999, Pärt et al. 2011, Ponchon et al. 2013, 2017). Moreover, individuals might use location cues provided by conspecific or even heterospecifics to select a breeding site (Thomson et al. 2003, Nocera et al. 2006, Betts et al. 2008, Thériault et al. 2012). Studies specifically examining prospecting behaviour are often limited to species whose life history facilitates its study, such as colonial species (Ponchon et al. 2017), species using nest boxes (Doligez et al. 2004) or species with territorial non- breeders (Pärt et al. 2011). Under these conditions it is possible to study non-breeding individuals interact and observe breeding conspecifics. While identifying prospecting individuals can be difficult, any movement occurring outside of an individual’s breeding territory may be considered prospecting, as they may be acquiring information that could affect their choice of a future breeding site (Reed 1993).
Access to potential breeding sites by prospectors during the post-breeding period may be critical to facilitate immigration into a specific area. Using hand-reared individuals, Löhrl (1959) found that Collared Flycatchers (Ficedula albicollis) returned to breed in the general area where they spent the period just preceding the first prebasic molt (Löhrl 1959), demonstrating the importance of this period in breeding site selection. Very little is known about the effects of fragmented habitat on the movements of individuals prospecting at this period. One of the few studies on this subject found that newly independent juvenile Crested Tits (Lophophanes cristatus) born in habitat fragments showed a delayed departure relative to individuals born in controls (plots within large tracts of habitat), suggesting a reluctance to cross forest fragment edges or to move through the matrix (Lens and Dhondt 1994).
Delayed departure from a natal territory may limit the time available to prospect. Additionally, it may limit the quality of information gathered, as some public information is season-sensitive (Reed et al. 1999). In general, very little is known about prospecting behaviour and even less in the context of fragmented landscapes.
In this study, we investigated whether landscape structure can influence the prospecting movements of a migratory songbird, the Ovenbird (Seiurus aurocapilla). More specifically, we determined whether landscape composition and configuration can influence capture rates of individuals moving through potential breeding habitat during the dispersal period occurring between the breeding season and fall migration. While the movements taking place between the breeding season and fall migration are poorly documented, they are believed to reflect, at least in part, foraging behaviour and predator avoidance (Vitz 2008, Streby and Andersen 2012). Hence, public information and environmental cues may be acquired incidentally as individuals venture onto potential breeding sites (Nocera et al. 2006). We assumed that movements taking place during the dispersal period could provide habitat cues, location cues, or public information to prospecting individuals. In turn, any effect of landscape structure on these movements could have implications for prospectors’ ability to collect such information.
Translocation studies suggest that Ovenbirds move more readily across the landscape as the proportion of mature forest cover increases (Bélisle et al. 2001, Gobeil and Villard 2002) and that they are reluctant to move across certain matrix types such as clearcuts (Bélisle and Desrochers 2002, Robichaud et al. 2002, Valente et al. 2019) and conifer plantations (Villard and Haché 2012). Yet, individuals tend to move faster when experimentally released in plantations than in deciduous forest, suggesting that they avoid crossing plantation edges (Geoffroy et al. 2019). Our objective is to understand how landscape structure influences prospecting movements during the post-fledging period. We predicted that the likelihood of capturing prospecting individuals in potential breeding habitat (i.e., mature deciduous stands)
will be negatively affected by the fragmentation of mature forest and by the proportion of conifer plantations and other matrix types relatively impermeable to movement, such as recent clearcuts (<5 years).
1.5 METHODS
Study Area
The study was conducted in a northern hardwood forest of northwestern New Brunswick, Canada, 20-35 km northeast of Edmundston (47° 31' 12N, 68° 8' 13W) (Figure 1). The study area is characterized by shade-tolerant deciduous forest codominated by sugar maple (Acer saccharum), American beech (Fagus grandifolia), and yellow birch (Betula alleghaniensis) on well-drained sites and red (Picea rubens) or black spruce (P. mariana) and balsam fir (Abies balsamea) in mesic or poorly drained sites. The region of the province is intensively managed for timber and features various types of partial cutting or shelterwood harvesting in deciduous stands, and clearcuts followed by spruce plantations in conifer- dominated stands. We selected 6 mature forest fragments large enough to encompass a 6 ha study plot. Plots were between 2 km and 22 km from one another. These forest fragments were characterized by mature deciduous forest dominated by sugar maple and yellow birch, with minimal signs of recent anthropogenic disturbance, except for the production of maple syrup (sugar bushes).
Focal Species
The Ovenbird is a neotropical migrant that breeds in the temperate deciduous or mixed forests of eastern North America (Porneluzi et al. 2011). It both nests and forages in deciduous leaf litter (Porneluzi et al. 2011). Males arrive first on the breeding grounds in the spring and quickly establish a territory (Thériault et al. 2012). In our study area, the breeding
season begins in mid-May and lasts for approximately 2 months. Adult Ovenbirds tend to be site faithful (Haché and Villard 2010), especially when they fledge young (Thériault et al. 2012), making the selection of their first territory particularly crucial.
Capture and Manipulations
Ovenbirds were captured under the animal care permit #16-13 at the University of Moncton. In each study plot, we then conducted constant-effort mist netting using four 12 m mist nets, from July 29, 2017 to August 9, 2017. Nets were arranged in pairs to create two 24 m net lanes, 100 m apart and at least 100 m from the forest edge. We used playbacks of conspecific vocalizations (songs) for the entire mist netting period. A speaker connected to an mp3 players was placed at eye level in vegetation adjacent to each net lane. Nets were opened at sunrise and closed 4 hours later, unless wind speed was too high, or precipitations lasted more than 30 min. Each plot was visited every 3 days, except when adverse weather caused delays (never more than 2 days). Sampling effort was equal in each plot, for a total of 30 site days and 480 net hours. Foreign individuals could be discriminated from local ones as we had previously banded 70-90% (n = 50) of all males holding territories in each study plot. Breeding females were not targeted for banding before the beginning of the study as they were much less responsive to playbacks. Additionally, we were unlikely to capture local hatch year individuals at the period we targeted as juvenile birds quickly leave their natal territory (Streby et al. 2011b, Vitz and Rodewald 2011) and, indeed, show very low site fidelity (Hann 1937). Therefore, the unbanded individuals we captured were deemed to be prospectors.
Each newly captured individual was banded using a numbered metal band and two rectrices (r3) were plucked, symmetrically, for ageing purposes. Only birds captured for the first time during the study were included in the analysis. Following Donovan and Stanley
(1995), we used the wear angle of the rectrix to assign individuals to age groups (<77.92ᵒ for HY/SY and >90.05ᵒ for AHY/ASY), where HY/SY are birds in their first year of life and AHY/ASY are those born before the previous year. Individuals were classified HY/SY if they still retained the rectrices grown in the nest and had yet to undergo their definitive prebasic molt. Individuals were classified as AHY/ASY if they were considered to be in their definitive plumage. Following Bayne and Hobson (2001a), we used the midpoint between the two angles (84ᵒ) to separate these age groups.
Landscape Characterization
Landscape metrics were quantified using the GIS database of Acadian Timber Corp. Two different landscape classifications were performed: one with seven land cover types, to emphasize the effects of silvicultural treatments on landscape permeability, and the other with four broader cover types, which were used to calculate fragmentation indices. The 7 cover types used in the initial characterization were: conifer plantation, recent clearcut (> 60% basal area removed, <5 years old), regenerating clearcut (> 60% basal area removed, 5- 20 years old), old clearcut (> 60% basal area removed, >20 years old), partial harvest (< 40 basal area removed), no treatment, and non-forest land cover types (mostly roads and bodies of water). These 7 categories were then grouped based of our knowledge of Ovenbird habitat use and previous translocation studies into 4 categories: low permeability to movement (conifer plantation, recent clearcut), high permeability (old clearcuts, partial harvest, no treatment), intermediate permeability (5-20 years old regenerating forest), and lastly, non- forest land cover (mostly roads and bodies of water).
After characterizing the landscape, we determined the proportion of each land class at three radii around the plots: 500 m, 1 km, and 2 km (Figure 1). These radii were selected to investigate the effects of landscape structure on movements at local and landscape scales around each plot. While warbler dispersal usually occurs in the distances of tens of kilometres
(Tittler et al. 2009), it is likely that landscape structure immediately adjacent to breeding habitat has the greatest influence on accessibility. Distances greater than 2 km were excluded because study plots were small and landscape homogeneity tended to increase with spatial scale in our study area. Proportions of each land cover type were calculated using a donut, which excluded a 100 radius around the point located at mid-distance between the two net pairs in each plot. The distance of 100 m was deemed to represent the distance at which individuals could hear the playback.
Fragmentation indices were also calculated at three scales matching those of the habitat proportions, but without excluding the first 100 m radius to avoid biasing calculations of fragmentation indices. Fragmentation metrics were calculated with FRAGSTATS (McGarigal and Eduard 2015). We quantified 8 landscape metrics: patch area, patch perimeter/area ratio, proximity index, mean proximity index, mean Euclidean nearest neighbour distance, connectivity index, class-level edge density and total edge density (see Table 1 for complete explanation of fragmentation metric calculations). Patch area was selected to determine whether prospection was more frequent in larger patches, whereas patch perimeter/area ratio was included to determine the influence of patch shape. The proximity index, mean proximity index, and mean Euclidean nearest neighbour distance were selected to characterize the structural isolation of focal patches (Wang et al. 2014). The connectivity index was used to quantify structural connectivity, whereas class (all patches of a same type) and landscape level edge densities were selected to detect potential edge effects.
Statistical Analyses
We modelled the influence of landscape structure on daily capture rate using general linear mixed models (GLMM) with a Poisson distribution. All GLMMs were fitted using the glmer function from the lme4 package (Bates et al. 2015)and performed using R (v. 3.4.3) (Core Team 2017). Daily capture rate was calculated as the number of new individuals
captured on a given day on a given plot as net hours remained constant on each day and plot. Given that we obtained repeated measures of capture rate on each plot, we considered study plot ID as a random variable (intercept). Explanatory variables were carefully selected to represent a range of hypotheses about potential landscape effects on prospecting movements, while restricting their number to account for sample size (Table 2). Conifer plantation and recent clearcut cover were selected as they have been shown to negatively affect Ovenbird movements (Bélisle et al. 2001, Villard and Haché 2012). Regenerating habitats were selected as they are often used by young Ovenbirds during the post fledging period (Streby et al. 2011b). Breeding habitat cover and patch area were tested as a highly positive correlation between capture rate and habitat amount could reflect the capture of local birds. Patch perimeter area ratio was chosen to reflect patch shape as a narrow patch may funnel birds into nets and increase capture rates. Patch proximity index, mean proximity index, mean Euclidean nearest neighbour distance and connectivity index were included as they reflect different measures of landscape connectivity and fragmentation. Class and landscape level edge densities were tested as Ovenbirds have previously been thought to be reluctant to cross edges (Bélisle et al. 2001), and thus we would expect lower capture rates in landscapes with high edge density. We also selected the interaction between percentage of permeable land cover and habitat structure, however high correlation between variables greatly limited possible combinations (see below).
We tested 17 a priori models selected to reflect both landscape composition and configuration. However, a few a priori models were removed due to collinearity (VIF) (see Table 2). All explanatory variables were standardized to a mean of zero and unit variance. We then used the Akaike information criterion corrected for small sample size (AICc) to evaluate the performance of individual variables and models using the function ICtab in R (Burnham and Anderson 2002). Oversidpersion was tested using the function overdisp_fun from the package lme4 (Bates et al. 2015). We ran 18 models (including a null model) at
each of the three spatial scales and selected those with a ΔAICc <2 from the best model at each spatial scale. The values reported in the results section are means ± CI95%, while marginal R2 are calculated using the function r.squaredGLMM and represented as percentages.
1.6 RESULTS
A total of 84 individuals were captured during the study. Of these, 7 were known territorial males that we had banded on the same site during the breeding season, and they were therefore excluded from subsequent analyses. Most (92%) of the unbanded individuals we captured were classified as HY/SY birds. Hatch year birds are considered most likely to be prospecting for a future breeding territory because they have not had the chance to collect information on breeding site quality yet (Reed 1993, Boulinier et al. 1996, Davis et al. 2017). However, as AHY individuals may also be prospecting, especially after a failed breeding attempt (Reed et al. 1999), we considered any individual found on a plot on which it did not previously hold a territory a putative prospector as they could have been acquiring information for future reproduction, i.e. prospecting (Ward 2005b).
Two models explaining daily capture rate of Ovenbird deemed to be prospectors were among the best (ΔAICc < 2) at the smallest spatial scale (see Table 2 for all ΔAICc values). The proportion of conifer plantations surrounding a plot was the top predictors of capture rate at this scale (Table 2; Model 1). The likelihood of capturing individuals decreased with the